A Layered Lattice Coding Scheme for a Class of Three User Gaussian Interference Channels

• The Han-Kobayashi achievable region was known to approximate capacity within one bit for two-user Gaussian interference channels, but that result did not extend beyond two users.Source: §I · p. 1
• For general fixed K-user interference channels, only one degree of freedom had been achieved, while the best outer bound was K/2.Known multi-user characterizations for time- or frequency-varying channels did not apply to fixed channels.Source: §I · p. 1
• Layered lattice coding improves rates over layered random coding and achieves more than one degree of freedom for a wide class of three-user channels.The paper states that its results can be extended to K-user interference channels.Source: §I · p. 1
• Lattice coding had previously been used for one-sided and very-strong interference channels and for AWGN capacity results.Source: §I · p. 1

• The model has three transmitter-receiver pairs, with independent message m_k from transmitter k intended for receiver k.Source: §II · p. 1
• The channel model uses gains h_jk from transmitter k to receiver j, and normalizes all direct channel gains to unity.Source: §II · p. 1
• The analysis restricts channel matrices to H1, a nontrivial class that includes the symmetric interference channel as a special case.Source: §II · p. 2

• Construction A generates a lattice from a linear (n, k) code over Z_p, and the paper uses balanced sets of such codes to form a set of lattices.Source: §III · p. 2
• For a point-to-point additive-noise channel, a suitable lattice and decoding strategy can make error probability arbitrarily small as lattice dimension increases.Source: §III · p. 2
• The proof of the lattice error-bound lemma is omitted, and the following lemma provides the single-user AWGN lattice-code existence result used in the paper.The paper notes that the omitted proof is described in.Source: §III · pp. 2–3
• A lattice-code existence result is stated for a single-user AWGN channel under a power constraint, using a Construction-A lattice, shift, and shaping region.Source: §III · p. 2

IV. SYMMETRIC IC: DEGREES OF FREEDOM

• The analysis uses a previously established very-strong-interference result for a symmetric three-user Gaussian IC, then applies it repeatedly across decoding stages.The cited lemma is stated for channel gain a, transmitter power P, and receiver noise variance σ^2.Source: §IV · p. 3
• Theorem 1 establishes more than one total degree of freedom for a symmetric three-user Gaussian IC over a wide range of channel parameters.The theorem assumes noise variance 1; time sharing gives one degree of freedom when 1/3 ≤ a^2 ≤ 2.Source: §IV · p. 3
• Each transmitter lattice-encodes N message sub-messages, assigning a rate and power to each layer and decoding through N stages.For a^2 ≥ 2, receivers decode interference before their intended message at each stage; for the other case, they decode the message first because interference-first constraints are more binding.Source: §IV · pp. 3–4
• The assigned powers make the very-strong-interference condition hold at each stage, enabling the per-stage rates used to form each user's total rate.Each sub-message rate is positive when a^2 ≥ 2, and the total power per transmitter is bounded by (2a^4 − a^2)^N.Source: §IV · p. 4
• The achievable total degrees of freedom tends to 3/2 as a approaches zero or infinity, and equals 3 at a = 0.Source: §IV · p. 5

V. COMPARING LATTICE CODING AND HAN KOBAYASHI CODING

• Han–Kobayashi coding uses Gaussian codebooks and splits each transmitter’s message into one private and three common parts.The common parts are intended for receiver 2, receiver 3, or both, as applicable.Source: §V · p. 5
• The layered lattice scheme significantly outperforms the three-user Han–Kobayashi extension despite potentially suboptimal identical power splitting.The authors note that different power-splitting schemes could possibly improve the lattice results.Source: §V · p. 6
• For a = 2.5, Fig. 3 compares the achievable lattice symmetric rate with the maximum Han–Kobayashi symmetric rate.The comparison is for the three-user symmetric Gaussian interference channel.Source: §V · p. 5, Figure 3 · p. 6
• In the three-user setting, the paper argues Han–Kobayashi coding is not optimal in degrees of freedom, unlike its within-one-bit result for two users.The authors suggest lattice-based interference alignment as a possible optimal strategy; decoding interference without decoding its messages removes MAC-type constraints on interfering-user rates.Source: §V · p. 6

VI. GENERALIZATIONS TO NON-SYMMETRIC CHANNELS

• For channels in H1, the section generalizes the very-strong-interference result to non-symmetric channels [18].The lemma’s rates hold when the channel gains satisfy one of its specified conditions.Source: §VI · p. 6
• Under the stated channel-gain conditions, each user can achieve rates within half a bit of their maximum possible rates.The result is for three-user channels whose matrix lies in H1 and whose gains satisfy one of the listed conditions.Source: §VI · p. 6
• The transmitters choose lattices so interference lattices align at each receiver, which decodes total interference before its own message.For receiver 1, the total interference lies in a lattice, and it is decoded in the presence of the desired signal and noise.Source: §VI · pp. 6–7
• When the channel gains satisfy (35), interference-decoding rate constraints are less binding than the constraints for decoding each user’s message.The paper then gives achievable user rates for this setting.Source: §VI · p. 7

A. Degree of Freedom Analysis

• Each transmitter splits its message into sub-messages, encodes them with lattice codebooks, and assigns power to each sub-message.The codebooks use lattices, shifts, and spherical shaping regions.Source: A. Degree of Freedom Analysis · p. 7
• At each decoding stage, the receiver decodes interference before its own sub-message, with powers chosen to satisfy a very-strong-interference condition.The condition makes interference-decoding rate constraints less binding than the constraint for decoding the intended message.Source: A. Degree of Freedom Analysis · p. 7
• The non-symmetric example’s total degrees of freedom are illustrated numerically because the analysis does not derive a closed-form expression.The example uses a1 = 2α, a2 = 3α, and a3 = 4α.Source: A. Degree of Freedom Analysis · pp. 7–8, Figure 4 · p. 8
• The analysis presents the same degree-of-freedom approach for other non-symmetric three-user channels in H1, with details omitted.Source: A. Degree of Freedom Analysis · pp. 7–8

VII. CONCLUSIONS

• The layered lattice coding approach achieves more than one degree of freedom for symmetric channels and yields an analytical expression in terms of channel gain a.Source: §VII · p. 8
• For symmetric interference channels, decoding interference rather than interfering messages provides significant rate benefits over the Han-Kobayashi strategy.Source: §VII · p. 8
• The paper extends the very-strong-interference result of [18] to a class of non-symmetric channels and uses it to generalize layered lattice coding.Source: §VII · p. 8